Data Scientist Nanodegree

Convolutional Neural Networks

Project: Write an Algorithm for a Dog Identification App

This notebook walks you through one of the most popular Udacity projects across machine learning and artificial intellegence nanodegree programs. The goal is to classify images of dogs according to their breed.

If you are looking for a more guided capstone project related to deep learning and convolutional neural networks, this might be just it. Notice that even if you follow the notebook to creating your classifier, you must still create a blog post or deploy an application to fulfill the requirements of the capstone project.

Also notice, you may be able to use only parts of this notebook (for example certain coding portions or the data) without completing all parts and still meet all requirements of the capstone project.

In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this IPython notebook.

Why We're Here

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 6: Write your Algorithm
  • Step 7: Test Your Algorithm

Step 0: Import Datasets

Import Dog Dataset

In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:

  • train_files, valid_files, test_files - numpy arrays containing file paths to images
  • train_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labels
  • dog_names - list of string-valued dog breed names for translating labels
In [1]:
from sklearn.datasets import load_files       
from keras.utils import np_utils
import numpy as np
from glob import glob

# define function to load train, test, and validation datasets
def load_dataset(path):
    data = load_files(path)
    dog_files = np.array(data['filenames'])
    dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
    return dog_files, dog_targets

# load train, test, and validation datasets
train_files, train_targets = load_dataset('../../../data/dog_images/train')
valid_files, valid_targets = load_dataset('../../../data/dog_images/valid')
test_files, test_targets = load_dataset('../../../data/dog_images/test')

# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("../../../data/dog_images/train/*/"))]

# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))

Using TensorFlow backend.

There are 133 total dog categories.
There are 8351 total dog images.

There are 6680 training dog images.
There are 835 validation dog images.
There are 836 test dog images.

Import Human Dataset

In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.

In [2]:
import random
random.seed(8675309)

# load filenames in shuffled human dataset
human_files = np.array(glob("../../../data/lfw/*/*"))
random.shuffle(human_files)

# print statistics about the dataset
print('There are %d total human images.' % len(human_files))

There are 13233 total human images.

Step 1: Detect Humans

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.

In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [3]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')

# load color (BGR) image
img = cv2.imread(human_files[10])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()

Number of faces detected: 1
In [4]:
## show 20 images with face detector;
fig = plt.figure(figsize=(25, 4))

for idx in np.arange(20):
    
    #load color (BGR) image
    img = cv2.imread(human_files[idx])
    # convert BGR image to grayscale
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    # find faces in image
    faces = face_cascade.detectMultiScale(gray)


    # get bounding box for each detected face
    for (x,y,w,h) in faces: cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
    # convert BGR image to RGB for plotting
    cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

    # display the image, along with bounding box
    ax = fig.add_subplot(2, 10, idx+1, xticks=[], yticks=[])
    plt.imshow(cv_rgb)

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [5]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer: 100% of human images with a detected face of the first 100 images in human_files; 11% of dog images with a detected face of the first 100 images in dog_files

In [6]:
human_files_short = human_files[:100]
dog_files_short = train_files[:100]
# Do NOT modify the code above this line.

## TODO: Test the performance of the face_detector algorithm 
## on the images in human_files_short and dog_files_short.
human_face = 0
for img in human_files_short:
    if(face_detector(img) == 1):
        human_face = human_face + 1
print("{} of human images with a detected face of the first 100 images in human_files".format(str(human_face)+'%'))

dog_face = 0
for img in dog_files_short:
    if(face_detector(img) == 1):
        dog_face = dog_face + 1
print("{} of dog images with a detected face of the first 100 images in dog_files".format(str(dog_face)+'%'))

100% of human images with a detected face of the first 100 images in human_files
11% of dog images with a detected face of the first 100 images in dog_files

Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?

Answer: Photo required user uploaded a clear view of face to build face detector algorithm, and it does not turn out as we wish, there is 11% mistake rate on dog imges (first 100 images). So I think build a CNN face detector might be a better apporach, CNN can draw from features of human faces, shape of human eyes, noses, mouth. For example, we all have two eyes, but it does not mean if there is only one eye had been captured on the photo makes a human face be a non human face, I could add more agumentation for the train data, like rotate, crop, etc. also can use pretrained CNN and alter it to suit my goal

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.

In [7]:
## (Optional) TODO: Report the performance of another  
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.

Step 2: Detect Dogs

In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.

In [8]:
from keras.applications.resnet50 import ResNet50

# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')

Downloading data from https://github.com/fchollet/deep-learning-models/releases/download/v0.2/resnet50_weights_tf_dim_ordering_tf_kernels.h5
102858752/102853048 [==============================] - 1s 0us/step

Pre-process the Data

When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape

$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$

where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.

The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape

$$ (1, 224, 224, 3). $$

The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape

$$ (\text{nb_samples}, 224, 224, 3). $$

Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!

In [9]:
from keras.preprocessing import image                  
from tqdm import tqdm

def path_to_tensor(img_path):
    # loads RGB image as PIL.Image.Image type
    img = image.load_img(img_path, target_size=(224, 224))
    # convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
    x = image.img_to_array(img)
    # convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
    return np.expand_dims(x, axis=0)

def paths_to_tensor(img_paths):
    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
    return np.vstack(list_of_tensors)
In [10]:
fig = plt.figure(figsize=(25, 4))

for idx in np.arange(20):
    
    #load color (BGR) image
    img = cv2.imread(train_files[idx])
    # convert BGR image to grayscale
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    
    # convert BGR image to RGB for plotting
    cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

    # display the image, along with bounding box
    ax = fig.add_subplot(2, 10, idx+1, xticks=[], yticks=[])
    plt.imshow(cv_rgb)
    ax.set_title(train_files[idx][35:50])

Making Predictions with ResNet-50

Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.

Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.

By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.

In [11]:
from keras.applications.resnet50 import preprocess_input, decode_predictions

def ResNet50_predict_labels(img_path):
    # returns prediction vector for image located at img_path
    img = preprocess_input(path_to_tensor(img_path))
    return np.argmax(ResNet50_model.predict(img))

Write a Dog Detector

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).

We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [12]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    prediction = ResNet50_predict_labels(img_path)
    return ((prediction <= 268) & (prediction >= 151))

(IMPLEMENTATION) Assess the Dog Detector

Question 3: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer:

  • 0% of dog face has been detected of the first 100 images in human_files
  • 100% of dog face has been detected of the first 100 images in dog_files
In [13]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.

TTrue = 0
for img in human_files_short:
    if(dog_detector(img) == 1):
        TTrue = TTrue + 1
print("{} of dog face has been detected of the first 100 images in human_files".format(str(TTrue)+'%'))

T_True = 0
for img in dog_files_short:
    if(dog_detector(img) == 1):
        T_True = T_True + 1
print("{} of dog face has been detected of the first 100 images in dog_files".format(str(T_True)+'%'))

0% of dog face has been detected of the first 100 images in human_files
100% of dog face has been detected of the first 100 images in dog_files

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador Black Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

Pre-process the Data

We rescale the images by dividing every pixel in every image by 255.

In [14]:
from PIL import ImageFile                            
ImageFile.LOAD_TRUNCATED_IMAGES = True                 

# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255

100%|██████████| 6680/6680 [01:09<00:00, 96.35it/s] 
100%|██████████| 835/835 [00:07<00:00, 108.89it/s]
100%|██████████| 836/836 [00:07<00:00, 110.01it/s]

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    model.summary()

We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:

Sample CNN

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.

Answer: I would like to build a Neural Nets with 3 convenultional layer, 3 max pooling layers to draw features from images, then global average pooling layer to reduce number of parameters,in addtional there are 1 dense layer act as classifier. All CNN layers are using 'relu' activation function except last one, last dense layer I use softmax function, also there are 20% dropout to avoid overfitting. In generela the structure of this CNN is: step 1-3 are features capturing, and 4 is the classfier:

  1. Input:(224 224 3) ---- Conv 1: 16 filters shape is: 3X3, padding 1 ----> out Conv1: (224 224 16) ---- Maxpooling 1: pool size 2X2 ----> out Maxpool 1: (112 112 16)

  2. In Conv2 (112 112 16) ---- Conv2: 32 filters, 3X3 shape, padding 1 ----> out Conv 2: (112 112 32) ----> Maxpooling 2: same pool size ----> out Maxpool 2: (56 56 32)

  3. In Conv 3 (56 56 32) ---- Conv3: 64 filters, padding 1 ----> outConv 3: (56 56 64 ) ----> Maxpooling 3: same pool size ----> out Maxpool 3: (28 28 64)

  4. Dropout rate 20%, global average pooling layer, finnaly into dense layer, out 133 classifier

In [15]:
from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential

model = Sequential()

### TODO: Define your architecture.
model.add(Conv2D(filters=16, kernel_size=3, padding='same', activation='relu', 
                        input_shape=(224, 224, 3)))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=32, kernel_size=3, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Conv2D(filters=64, kernel_size=3, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.2))
model.add(GlobalAveragePooling2D()) ## global avg pooling
model.add(Dense(133, activation='softmax'))

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 224, 224, 16)      448       
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 112, 112, 16)      0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 112, 112, 32)      4640      
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 56, 56, 32)        0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 56, 56, 64)        18496     
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 28, 28, 64)        0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 28, 28, 64)        0         
_________________________________________________________________
global_average_pooling2d_1 ( (None, 64)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 133)               8645      
=================================================================
Total params: 32,229
Trainable params: 32,229
Non-trainable params: 0
_________________________________________________________________

Compile the Model

In [16]:
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [17]:
from keras.callbacks import ModelCheckpoint  

### TODO: specify the number of epochs that you would like to use to train the model.

epochs = 10  #as start to see if the accuracy metric decrease or not

### Do NOT modify the code below this line.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5', 
                               verbose=1, save_best_only=True)

model.fit(train_tensors, train_targets, 
          validation_data=(valid_tensors, valid_targets),
          epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)

Train on 6680 samples, validate on 835 samples
Epoch 1/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.8826 - acc: 0.0104Epoch 00001: val_loss improved from inf to 4.86773, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.8826 - acc: 0.0103 - val_loss: 4.8677 - val_acc: 0.0132
Epoch 2/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.8362 - acc: 0.0146Epoch 00002: val_loss improved from 4.86773 to 4.82196, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.8361 - acc: 0.0145 - val_loss: 4.8220 - val_acc: 0.0216
Epoch 3/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.7968 - acc: 0.0174Epoch 00003: val_loss improved from 4.82196 to 4.81959, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.7966 - acc: 0.0175 - val_loss: 4.8196 - val_acc: 0.0192
Epoch 4/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.7644 - acc: 0.0209Epoch 00004: val_loss improved from 4.81959 to 4.76350, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.7645 - acc: 0.0208 - val_loss: 4.7635 - val_acc: 0.0192
Epoch 5/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.7308 - acc: 0.0243Epoch 00005: val_loss improved from 4.76350 to 4.75444, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.7311 - acc: 0.0243 - val_loss: 4.7544 - val_acc: 0.0240
Epoch 6/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.7017 - acc: 0.0258Epoch 00006: val_loss improved from 4.75444 to 4.72225, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.7025 - acc: 0.0257 - val_loss: 4.7223 - val_acc: 0.0299
Epoch 7/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.6653 - acc: 0.0318Epoch 00007: val_loss improved from 4.72225 to 4.68366, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.6657 - acc: 0.0317 - val_loss: 4.6837 - val_acc: 0.0275
Epoch 8/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.6290 - acc: 0.0338Epoch 00008: val_loss did not improve
6680/6680 [==============================] - 23s 3ms/step - loss: 4.6290 - acc: 0.0337 - val_loss: 4.7400 - val_acc: 0.0263
Epoch 9/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.5980 - acc: 0.0368Epoch 00009: val_loss improved from 4.68366 to 4.63837, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.5978 - acc: 0.0371 - val_loss: 4.6384 - val_acc: 0.0347
Epoch 10/10
6660/6680 [============================>.] - ETA: 0s - loss: 4.5589 - acc: 0.0413Epoch 00010: val_loss improved from 4.63837 to 4.62497, saving model to saved_models/weights.best.from_scratch.hdf5
6680/6680 [==============================] - 23s 3ms/step - loss: 4.5589 - acc: 0.0412 - val_loss: 4.6250 - val_acc: 0.0371
Out[17]:
<keras.callbacks.History at 0x7f10004d2d30>

Load the Model with the Best Validation Loss

In [18]:
model.load_weights('saved_models/weights.best.from_scratch.hdf5')

Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.

In [19]:
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]

# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 3.9474%

Step 4: Use a CNN to Classify Dog Breeds

To reduce training time without sacrificing accuracy, we show you how to train a CNN using transfer learning. In the following step, you will get a chance to use transfer learning to train your own CNN.

Obtain Bottleneck Features

In [20]:
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']

Model Architecture

The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.

In [21]:
VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))

VGG16_model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_2 ( (None, 512)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 133)               68229     
=================================================================
Total params: 68,229
Trainable params: 68,229
Non-trainable params: 0
_________________________________________________________________

Compile the Model

In [22]:
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

Train the Model

In [23]:
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5', 
                               verbose=1, save_best_only=True)

VGG16_model.fit(train_VGG16, train_targets, 
          validation_data=(valid_VGG16, valid_targets),
          epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)

Train on 6680 samples, validate on 835 samples
Epoch 1/20
6500/6680 [============================>.] - ETA: 0s - loss: 12.3242 - acc: 0.1242Epoch 00001: val_loss improved from inf to 10.59191, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 290us/step - loss: 12.2857 - acc: 0.1265 - val_loss: 10.5919 - val_acc: 0.2168
Epoch 2/20
6580/6680 [============================>.] - ETA: 0s - loss: 9.9417 - acc: 0.2874Epoch 00002: val_loss improved from 10.59191 to 9.87100, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 245us/step - loss: 9.9659 - acc: 0.2853 - val_loss: 9.8710 - val_acc: 0.2778
Epoch 3/20
6620/6680 [============================>.] - ETA: 0s - loss: 9.3857 - acc: 0.3509Epoch 00003: val_loss improved from 9.87100 to 9.55697, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 9.3765 - acc: 0.3512 - val_loss: 9.5570 - val_acc: 0.3126
Epoch 4/20
6600/6680 [============================>.] - ETA: 0s - loss: 9.1262 - acc: 0.3856Epoch 00004: val_loss improved from 9.55697 to 9.36114, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 245us/step - loss: 9.1555 - acc: 0.3837 - val_loss: 9.3611 - val_acc: 0.3377
Epoch 5/20
6520/6680 [============================>.] - ETA: 0s - loss: 8.9273 - acc: 0.4097Epoch 00005: val_loss improved from 9.36114 to 9.32527, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 8.9336 - acc: 0.4094 - val_loss: 9.3253 - val_acc: 0.3461
Epoch 6/20
6580/6680 [============================>.] - ETA: 0s - loss: 8.8022 - acc: 0.4278Epoch 00006: val_loss improved from 9.32527 to 9.11473, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 8.8006 - acc: 0.4274 - val_loss: 9.1147 - val_acc: 0.3725
Epoch 7/20
6640/6680 [============================>.] - ETA: 0s - loss: 8.6872 - acc: 0.4354Epoch 00007: val_loss improved from 9.11473 to 8.98008, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 245us/step - loss: 8.6784 - acc: 0.4356 - val_loss: 8.9801 - val_acc: 0.3557
Epoch 8/20
6560/6680 [============================>.] - ETA: 0s - loss: 8.3704 - acc: 0.4537Epoch 00008: val_loss improved from 8.98008 to 8.72924, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 8.3783 - acc: 0.4534 - val_loss: 8.7292 - val_acc: 0.3844
Epoch 9/20
6580/6680 [============================>.] - ETA: 0s - loss: 8.2478 - acc: 0.4673Epoch 00009: val_loss improved from 8.72924 to 8.67915, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 247us/step - loss: 8.2430 - acc: 0.4675 - val_loss: 8.6791 - val_acc: 0.3832
Epoch 10/20
6460/6680 [============================>.] - ETA: 0s - loss: 8.1817 - acc: 0.4780Epoch 00010: val_loss improved from 8.67915 to 8.64823, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 244us/step - loss: 8.1969 - acc: 0.4771 - val_loss: 8.6482 - val_acc: 0.4036
Epoch 11/20
6440/6680 [===========================>..] - ETA: 0s - loss: 8.1305 - acc: 0.4783Epoch 00011: val_loss improved from 8.64823 to 8.44737, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 244us/step - loss: 8.0948 - acc: 0.4805 - val_loss: 8.4474 - val_acc: 0.4036
Epoch 12/20
6640/6680 [============================>.] - ETA: 0s - loss: 7.8728 - acc: 0.4922Epoch 00012: val_loss improved from 8.44737 to 8.26828, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 245us/step - loss: 7.8644 - acc: 0.4928 - val_loss: 8.2683 - val_acc: 0.4024
Epoch 13/20
6600/6680 [============================>.] - ETA: 0s - loss: 7.6379 - acc: 0.5111Epoch 00013: val_loss improved from 8.26828 to 8.16352, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 247us/step - loss: 7.6354 - acc: 0.5109 - val_loss: 8.1635 - val_acc: 0.4168
Epoch 14/20
6520/6680 [============================>.] - ETA: 0s - loss: 7.5665 - acc: 0.5170Epoch 00014: val_loss improved from 8.16352 to 8.05709, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 7.5514 - acc: 0.5180 - val_loss: 8.0571 - val_acc: 0.4311
Epoch 15/20
6440/6680 [===========================>..] - ETA: 0s - loss: 7.5399 - acc: 0.5252Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 2s 243us/step - loss: 7.5212 - acc: 0.5262 - val_loss: 8.1509 - val_acc: 0.4275
Epoch 16/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.4692 - acc: 0.5292Epoch 00016: val_loss improved from 8.05709 to 8.00570, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 245us/step - loss: 7.4794 - acc: 0.5286 - val_loss: 8.0057 - val_acc: 0.4323
Epoch 17/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.3968 - acc: 0.5358Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 2s 246us/step - loss: 7.4005 - acc: 0.5356 - val_loss: 8.0224 - val_acc: 0.4299
Epoch 18/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.3911 - acc: 0.5381Epoch 00018: val_loss improved from 8.00570 to 7.96163, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 7.3854 - acc: 0.5383 - val_loss: 7.9616 - val_acc: 0.4443
Epoch 19/20
6540/6680 [============================>.] - ETA: 0s - loss: 7.3843 - acc: 0.5391Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 2s 244us/step - loss: 7.3841 - acc: 0.5392 - val_loss: 7.9822 - val_acc: 0.4407
Epoch 20/20
6620/6680 [============================>.] - ETA: 0s - loss: 7.3794 - acc: 0.5406Epoch 00020: val_loss improved from 7.96163 to 7.95103, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s 246us/step - loss: 7.3807 - acc: 0.5406 - val_loss: 7.9510 - val_acc: 0.4491
Out[23]:
<keras.callbacks.History at 0x7f0f3d986470>

Load the Model with the Best Validation Loss

In [24]:
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')

Test the Model

Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.

In [25]:
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 44.6172%

Predict Dog Breed with the Model

In [26]:
from extract_bottleneck_features import *

def VGG16_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = VGG16_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:

The files are encoded as such:

Dog{network}Data.npz

where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.

(IMPLEMENTATION) Obtain Bottleneck Features

In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:

bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
In [27]:
### TODO: Obtain bottleneck features from another pre-trained CNN.
bottleneck_features = np.load('bottleneck_features/DogResnet50Data.npz')
train_Resnet50 = bottleneck_features['train']
valid_Resnet50 = bottleneck_features['valid']
test_Resnet50 = bottleneck_features['test']

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    <your model's name>.summary()

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer: I use almost everything from the pretrained CNN ResNet50, weights and architecture. Because in our case, the dataset is relatively really small , only around 10k images and only 133 dog breed classes in total. However RestNet50 got trained over millions of images and over 1000 classes, so use ResNet50 is a overkill for our case, use it as a fixed feature extractor, where the last convolutional output of ResNet is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the layer contains one node for each dog category (133 classes in total) and is equipped with a softmax activation function

In [45]:
### TODO: Define your architecture.
Resnet50_model = Sequential()
Resnet50_model.add(GlobalAveragePooling2D(input_shape=train_Resnet50.shape[1:]))
Resnet50_model.add(Dense(133, activation='softmax'))

Resnet50_model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_7 ( (None, 2048)              0         
_________________________________________________________________
dense_10 (Dense)             (None, 133)               272517    
=================================================================
Total params: 272,517
Trainable params: 272,517
Non-trainable params: 0
_________________________________________________________________

(IMPLEMENTATION) Compile the Model

In [46]:
### TODO: Compile the model.
Resnet50_model.compile(loss='categorical_crossentropy', optimizer='SGD', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [47]:
### TODO: Train the model.
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.Resnet50.hdf5', 
                               verbose=1, save_best_only=True)

Resnet50_model.fit(train_Resnet50, train_targets, 
          validation_data=(valid_Resnet50, valid_targets),
          epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)

Train on 6680 samples, validate on 835 samples
Epoch 1/20
6580/6680 [============================>.] - ETA: 0s - loss: 3.0101 - acc: 0.3924Epoch 00001: val_loss improved from inf to 1.72501, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 2s 261us/step - loss: 2.9902 - acc: 0.3970 - val_loss: 1.7250 - val_acc: 0.6455
Epoch 2/20
6640/6680 [============================>.] - ETA: 0s - loss: 1.2172 - acc: 0.7482Epoch 00002: val_loss improved from 1.72501 to 1.12682, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 212us/step - loss: 1.2155 - acc: 0.7485 - val_loss: 1.1268 - val_acc: 0.7377
Epoch 3/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.8107 - acc: 0.8292Epoch 00003: val_loss improved from 1.12682 to 0.90793, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 213us/step - loss: 0.8088 - acc: 0.8296 - val_loss: 0.9079 - val_acc: 0.7808
Epoch 4/20
6520/6680 [============================>.] - ETA: 0s - loss: 0.6213 - acc: 0.8699Epoch 00004: val_loss improved from 0.90793 to 0.80803, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 214us/step - loss: 0.6207 - acc: 0.8701 - val_loss: 0.8080 - val_acc: 0.7856
Epoch 5/20
6560/6680 [============================>.] - ETA: 0s - loss: 0.5123 - acc: 0.9002Epoch 00005: val_loss improved from 0.80803 to 0.74128, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 213us/step - loss: 0.5111 - acc: 0.8996 - val_loss: 0.7413 - val_acc: 0.8096
Epoch 6/20
6460/6680 [============================>.] - ETA: 0s - loss: 0.4362 - acc: 0.9175Epoch 00006: val_loss improved from 0.74128 to 0.69812, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 210us/step - loss: 0.4361 - acc: 0.9178 - val_loss: 0.6981 - val_acc: 0.8084
Epoch 7/20
6420/6680 [===========================>..] - ETA: 0s - loss: 0.3787 - acc: 0.9333Epoch 00007: val_loss improved from 0.69812 to 0.67020, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.3786 - acc: 0.9326 - val_loss: 0.6702 - val_acc: 0.8216
Epoch 8/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.3356 - acc: 0.9477Epoch 00008: val_loss improved from 0.67020 to 0.64709, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.3364 - acc: 0.9473 - val_loss: 0.6471 - val_acc: 0.8204
Epoch 9/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.3008 - acc: 0.9524Epoch 00009: val_loss improved from 0.64709 to 0.62729, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.3013 - acc: 0.9521 - val_loss: 0.6273 - val_acc: 0.8156
Epoch 10/20
6660/6680 [============================>.] - ETA: 0s - loss: 0.2726 - acc: 0.9646Epoch 00010: val_loss improved from 0.62729 to 0.61401, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.2728 - acc: 0.9642 - val_loss: 0.6140 - val_acc: 0.8275
Epoch 11/20
6440/6680 [===========================>..] - ETA: 0s - loss: 0.2480 - acc: 0.9688Epoch 00011: val_loss improved from 0.61401 to 0.60270, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 210us/step - loss: 0.2483 - acc: 0.9687 - val_loss: 0.6027 - val_acc: 0.8275
Epoch 12/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.2279 - acc: 0.9754Epoch 00012: val_loss improved from 0.60270 to 0.59608, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.2283 - acc: 0.9753 - val_loss: 0.5961 - val_acc: 0.8275
Epoch 13/20
6640/6680 [============================>.] - ETA: 0s - loss: 0.2115 - acc: 0.9789Epoch 00013: val_loss improved from 0.59608 to 0.57793, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.2111 - acc: 0.9790 - val_loss: 0.5779 - val_acc: 0.8347
Epoch 14/20
6580/6680 [============================>.] - ETA: 0s - loss: 0.1962 - acc: 0.9815Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 1s 209us/step - loss: 0.1965 - acc: 0.9816 - val_loss: 0.5783 - val_acc: 0.8311
Epoch 15/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.1829 - acc: 0.9832Epoch 00015: val_loss improved from 0.57793 to 0.56954, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 213us/step - loss: 0.1825 - acc: 0.9834 - val_loss: 0.5695 - val_acc: 0.8359
Epoch 16/20
6480/6680 [============================>.] - ETA: 0s - loss: 0.1713 - acc: 0.9844Epoch 00016: val_loss improved from 0.56954 to 0.56203, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 210us/step - loss: 0.1715 - acc: 0.9841 - val_loss: 0.5620 - val_acc: 0.8311
Epoch 17/20
6580/6680 [============================>.] - ETA: 0s - loss: 0.1612 - acc: 0.9866Epoch 00017: val_loss improved from 0.56203 to 0.55769, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 212us/step - loss: 0.1611 - acc: 0.9867 - val_loss: 0.5577 - val_acc: 0.8371
Epoch 18/20
6580/6680 [============================>.] - ETA: 0s - loss: 0.1512 - acc: 0.9889Epoch 00018: val_loss improved from 0.55769 to 0.55103, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 212us/step - loss: 0.1514 - acc: 0.9888 - val_loss: 0.5510 - val_acc: 0.8359
Epoch 19/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.1431 - acc: 0.9909Epoch 00019: val_loss improved from 0.55103 to 0.55096, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 211us/step - loss: 0.1430 - acc: 0.9910 - val_loss: 0.5510 - val_acc: 0.8347
Epoch 20/20
6540/6680 [============================>.] - ETA: 0s - loss: 0.1362 - acc: 0.9914Epoch 00020: val_loss improved from 0.55096 to 0.54456, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s 213us/step - loss: 0.1355 - acc: 0.9916 - val_loss: 0.5446 - val_acc: 0.8395
Out[47]:
<keras.callbacks.History at 0x7f0f54fa0a20>

(IMPLEMENTATION) Load the Model with the Best Validation Loss

In [48]:
### TODO: Load the model weights with the best validation loss.
Resnet50_model.load_weights('saved_models/weights.best.Resnet50.hdf5')

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.

In [49]:
### TODO: Calculate classification accuracy on the test dataset.
# get index of predicted dog breed for each image in test set
Resnet50_predictions = [np.argmax(Resnet50_model.predict(np.expand_dims(feature, axis=0))) for feature in test_Resnet50]

# report test accuracy
test_accuracy = 100*np.sum(np.array(Resnet50_predictions)==np.argmax(test_targets, axis=1))/len(Resnet50_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 85.0478%

(IMPLEMENTATION) Predict Dog Breed with the Model

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.

Similar to the analogous function in Step 5, your function should have three steps:

  1. Extract the bottleneck features corresponding to the chosen CNN model.
  2. Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
  3. Use the dog_names array defined in Step 0 of this notebook to return the corresponding breed.

The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function

extract_{network}

where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.

In [50]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.

from extract_bottleneck_features import *

def Resnet50_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_Resnet50(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = Resnet50_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 6: Write your Algorithm

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.

A sample image and output for our algorithm is provided below, but feel free to design your own user experience!

Sample Human Output

This photo looks like an Afghan Hound.

(IMPLEMENTATION) Write your Algorithm

In [66]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.
import matplotlib.pyplot as plt
import matplotlib.image as mpimg


def human_or_dog(img_path):
    img = mpimg.imread(img_path)
    _ = plt.imshow(img)
    
    if dog_detector(img_path):
        print('Dog breed: "{}"'.format(Resnet50_predict_breed(img_path)[15:]))
        
    elif face_detector(img_path):
        print('Human resemble "{}"'.format(Resnet50_predict_breed(img_path)[15:]))
    
    else:
        print('Error')
    
    plt.show()
    return
In [68]:
import os

def get_random_images(home_path, num_imgs):
    '''
    get 10 random dog images from test file
    To test how our model perform 
    '''
    files = []
    while len(files)<num_imgs:
        folder = random.choice(os.listdir(home_path))
        newpath = home_path+'/'+folder
        file = random.choice(os.listdir(newpath))
        finalpath = newpath+'/'+file
        if os.path.isfile(finalpath)==True and finalpath[-3:]=='jpg':
            files.append(finalpath)
    return files

#get some random images from the TEST data which the network was not trained on.
files = get_random_images('../../../data/dog_images/test',10)
for file in files:
    print('\n'+file[30:60])
    human_or_dog(file)

087.Irish_terrier/Irish_terrie
Dog breed: "Norfolk_terrier"

022.Belgian_tervuren/Belgian_t
Dog breed: "Belgian_tervuren"

100.Lowchen/Lowchen_06685.jpg
Dog breed: "Havanese"

036.Briard/Briard_02539.jpg
Dog breed: "Briard"

095.Kuvasz/Kuvasz_06409.jpg
Dog breed: "Great_pyrenees"

035.Boykin_spaniel/Boykin_span
Dog breed: "Boykin_spaniel"

074.Giant_schnauzer/Giant_schn
Dog breed: "Giant_schnauzer"

019.Bedlington_terrier/Bedling
Dog breed: "Bedlington_terrier"

087.Irish_terrier/Irish_terrie
Dog breed: "Irish_terrier"

027.Bloodhound/Bloodhound_0190
Dog breed: "Bloodhound"

Step 7: Test Your Algorithm

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

Answer: Overall, it performs well to me. Test Accuracy is around 84 percente. I used both 2 images from test data and I uploaded 13 dog, human images to Dog_img folder to run our test. Fot the images from our test data, the model performs excellent, got both labels right. For the images I selected online, it perfroms ok. There are few reason for it, for the images I chose, other than regular dog and human face, I also select few dog face with style transfered, like colour block and cartoon style dog, also I chose one with a girl hugging a dog. The model made mistakes on these few tricky images, there are few reasons for it.

  1. For the different style dogs, model cant do a good job. There is a room for improvment, thats is add more varieties to our datatset, like transfered style dog images, this is not like agumentation. What I think can be done is we could select different style of dogs online then add to our dataset, or we could do style transfer for dog data. However, generally sepeaking, to improve our model performace, we could add agumentaion to our training set, like rorate, crop, etc. More details can be found here

  2. Train on more epochs, maybe not now, we already see validation loss stopped deceasing significantly within 20 epochs, so keep training is not a good option now to improve our model. I think we could add addtional one dense layer after the global average pooling layer, its quite a big jump from the oringinal global avg pooling layer 2048 nodes to 133 classes (out), so if we could add one additional layer in between, ex model.add(Dense(512, activation ='relu')) then to next dense layer, this is an idea, maybe 2 additional dense layers 1024 --> 512 --> 133 works better. There are many options we could do here to see how the performace changes.

  3. Hyperparameter tunning, I switched optimizer from 'rmsprop' to 'SGD'(stochastic gradient decent ), learning rate, etc. There is a Tuner in Keras. Details can be found here. There is no global set of hyperparamters, so we could try more options and find out what suits our goal the best.

In [58]:
import os
# Dog_img folder has 15 images I collected from internet, so lets test all out

directory = r'Dog_img'
for filename in os.listdir(directory):
    if filename.endswith(".jpg"):
         human_or_dog(os.path.join(directory,filename))

Dog breed: "Border_collie"

Human resemble "Norwegian_lundehund"

Dog breed: "French_bulldog"

Error

Dog breed: "Afghan_hound"

Error

Dog breed: "Chow_chow"

Dog breed: "French_bulldog"

Dog breed: "Boxer"

Dog breed: "Bulldog"

Human resemble "Chinese_crested"

Human resemble "Bullmastiff"

Error

Dog breed: "American_foxhound"

Human resemble "Chihuahua"
In [ ]: